TheClimateChanger

I’ve read through the Christy & Spencer paper and I think it’s worth discussing, but a lot of the conclusions depend heavily on how the analysis is set up. First, it’s U.S.-only, which already limits how broadly you can interpret it. The U.S. is a small, noisy region with a lot of land-use influence, and it’s not necessarily representative of global behavior. Second, they lean on minimally adjusted station data, which is not a neutral choice. One big issue there is time of observation bias (TOBs)—earlier stations often used afternoon observation times, which can effectively double-count hot days across two calendar days. That tends to inflate extreme heat in the early part of the record relative to today, when observations are mostly taken in the morning. If you don’t correct for that, you’re giving early decades a built-in advantage in heat extremes. On top of that, the way “extremes” are counted matters a lot. If you’re relying on record-setting events and not handling ties properly (or only counting first occurrences), later decades are inherently disadvantaged. Early in the record, everything is a “new record.” Later on, even in a warmer climate, you’re more likely to tie or narrowly exceed previous values. If ties aren’t treated equally, you can manufacture a downward trend in “new extremes” even if the underlying distribution is shifting warmer. Another big limitation is that the analysis focuses almost entirely on summer high extremes and winter low extremes, which is a pretty narrow slice of the climate system. It doesn’t really tell you anything about extreme warmth in winter, spring, or fall—which is where a lot of the warming signal actually shows up—or about the broader shift in temperatures. Summer daytime highs (Tmax) are one of the least responsive variables in many parts of the U.S., especially in the East. And this is where land-use changes become critical. The eastern and central U.S. have undergone significant reforestation over the past century after being heavily clear-cut, which increases evapotranspiration and tends to suppress daytime highs. At the same time, the Midwest has seen an explosion in corn and soybean agriculture, and those crops transpire enormous amounts of water—far more than the historical landscape. That effectively adds a massive, artificial cooling mechanism during the growing season. So you’ve got large parts of the country where land surface changes are actively dampening summer heat extremes, even as the broader climate warms. Meanwhile, more arid regions in the West—where you don’t have that same vegetation-driven cooling—show increasing heat extremes, which is more in line with expectations. On top of that, the 1930s Dust Bowl looms large in any long-term U.S. extremes analysis. That period featured extreme heat driven in large part by land degradation and drought—largely human-influenced surface conditions. If your methodology already favors early extremes (via TOBs and record-counting asymmetry), and you anchor the dataset with the most extreme decade in the record for partly non-climatic reasons, you’re stacking the deck toward showing a long-term decline. There are also the usual issues with long-term station data—station moves, instrument changes, siting differences, etc. Homogenization is meant to deal with those. Skipping it doesn’t remove bias, it just means you’re accepting a different set of biases, many of which tend to inflate earlier extremes relative to modern ones. To be fair, the paper is very likely correct that cold extremes are declining—that’s one of the most consistent and physically well-understood signals in the observational record. It’s also true that regional trends (especially in the U.S.) can look different due to land-use effects. But the headline claim about declining heat extremes is much more questionable and looks heavily dependent on methodology. At a minimum, this analysis is incomplete—it’s looking at a narrow set of metrics that are least likely to show strong warming signals and then generalizing from them. If you actually want to understand how extremes are changing, you need to look at the full temperature distribution, across all seasons, with methods that account for known observational biases and major land surface changes.

With spring about halfway through, currently warmest April to date and 3rd warmest season to date [warmest in the KPIT era]. April is running about 1°F above 2010, and spring to date 0.5°F above 2012. Honestly, I'm more surprised it has been the 2nd wettest spring to date. I know it has been fairly wet, but I didn't know it had been that wet (we really haven't had any significant, widespread flooding or anything). 3rd warmest spring:

Incredible! Downloaded data from SERCC (observations from March 1-April 14, forecast values through April 19) reveals nearly 80 long-term threaded stations are in the midst of their warmest spring on record, calculated by average daily high temperature. Led by Huntington, West Virginia, where the first 50 days of spring has seen a mean high temperature of 72.8F, an astounding 12.1F above the 1991-2020 mean. Again, that's a 50-day average!

Here's where we could be sitting on Friday, with current forecast values. Hottest spring on record to date: Columbus, OH; Cincinnati, OH; all of Kentucky; Evansville, IN; and Springfield, MO. Top 2-5: Kansas City, MO; Des Moines, IA; Waterloo, IA; Moline, IL; Rockford, IL; Peoria, IL; Springfield, IL; Columbia, MO; Indianapolis, IN; South Bend, IN; Fort Wayne, IN; Toledo, OH; Dayton, OH; Mansfield, OH; Cleveland, OH; Akron/Canton, OH; Youngstown, OH. Top 6-10: Detroit, MI; Flint, MI; Lansing, MI; Grand Rapids, MI; Chicago, IL; Milwaukee, IL; Sioux City, IA; Dubuque, IA; La Crosse, WI; Minneapolis, MN; Rochester, MN. Really feel for those in the far north who apparently have missed out. Dare I say it - best climo ever?

Best spring in the history of America.

Hmm, maybe because the old measurements are biased low? Snowfall measurement: a flaky history | NCAR & UCAR News

Unreal. This may go into the annals as the most pleasant spring in the history of the City of Pittsburgh. Already up to 29 days at or above 60F (T-most), and 13 days at or above 70F (T-3rd most). Tomorrow will be close to 60F, but the ensuing 6 days are 70s and 80s. Count on those two should be 35/36 & 19, respectively, by next Friday, with even warmer temperatures possible next weekend.

I'm sure the observer in Toronto was lying about how much snow used to fall as well.

Are we sure these correlations apply in this new regime? If we look at April 1997, it was 5th coldest on record, but this year seems destined for back to back record warm months. The NBM numbers would bring us up to about +5.5F (1991-2020 climo) by the 18th/19th timeframe, whereas the current record is +3.4 in 2006.

Looks like good, sound records to me. The snowfall and liquid equivalents make sense.

What motive would they have to lie about how much snow fell? Newark is not the same climate as Detroit. It could have been colder and drier in Detroit than present.

This is unbelievable:

Ever since they swapped in the new temperature sensors, PIT's lows are usually no more than about 1F warmer than mine and I'm in the woods. It used to be 2-3, maybe 4F, warmer, but not recently.

Also, the afternoon/early evening observation time common at the cooperative sites in the earlier decades often tacks on an additional day at or above a threshold temperature to warm spells, since the set temperature for the following day would be only a little bit colder than the daily high.

In any event, this is going to be the hottest March on record at Death Valley National Park by a HUGE margin (pop: zero, wilderness station).

Not a very good example. FOK was more of a microclimate, if it's even a legit measurement. It was lower than any other location, regardless of rural/urban location, for hundreds of miles. Nothing lower until deep into New England. It was like 5F colder than even the coldest personal weather stations on Long Island.